Computing Error Measures for Migration Distance Estimates in Historical Linked Data Sets

Rebecca J. Vick, University of Minnesota
Sula Sarkar, University of Minnesota

The IPUMS group at the Minnesota Population Center has constructed twenty-one linked samples of United States census data from the 19th and early 20th centuries. Included in all twenty-one linked samples is an IPUMS-constructed variable called MILEMIG, which contains the number of miles a linked individual migrated between census years. This distance measure provides a unique perspective on migrant behavior during this great migration period in United States history. However, the precision of the MILEMIG data is on the county level thus should be used with caution. Providing a measure of error for MILEMIG would counteract researchers having a false sense of precision when using it in their research. The 95 percentile maximum mile error for New England is 46 miles. This regional error measure will help researchers use the variable for migration studies appropriately within their region(s) of interest.